An approach for detecting deviations in daily routine for long-term behavior analysis

Abstract

Rendering and offering adequate reminder services in a situation-aware, proactive manner and providing information for diagnosis support is a major issue for Ambient Assisted Living systems when it comes to dealing with persons suffering from mild dementia. One great challenge therefore is to reliably recognize and assess the long-term behavior of assisted persons. In the context of diagnosis support for caregivers or practitioners, deviations in the daily routine of a person with mild dementia might be an indicator of a deterioration of the affected person's cognitive condition. Based on this information, adequate help can be provided. We developed an approach to processing information regarding the modeling of daily routines and a comparison to previous days. Our solution can be seen as a combination of three approaches: a cosinor analysis based on the theory of circadian rhythms as a special representative of regression analysis, a histogram-based approach based on movement data, and a probabilistic model of behavior (PMB) based on the person's activities of daily living (ADL).